Shopify customer support triage workflow
Classify incoming customer emails with AI and route urgent issues to the right person fast.
This is a planning recipe. Check platform-specific steps before going live.
Who this workflow is for
E-commerce store owner — especially Shopify store.
What this workflow automates
Cut response time on Shopify support by letting AI tag each email and route urgent ones to Slack.
Tools required
Supported platforms
Visual workflow
Step-by-step setup
- 1New support email · Gmail
Watches your support inbox for new messages.
- • Connect Gmail
- • Filter to support label or address
- • Test with one real email
- 2Look up customer order · Shopify
Find the customer's most recent order using the sender email.
- • Search orders by email
- • Pull last order id, status, fulfillment
- 3Classify the request · OpenAI
Use AI to classify intent and urgency.
- • Use gpt-4o-mini or similar
- • Pass the email body and order JSON
- • Parse the JSON response
- 4Route by category · Router
Send urgent issues to Slack, refunds to the refund queue, others to a normal label.
- • Add router branches per category
- • Add an extra branch for urgency = high
- 5Alert team for urgent issues · Slack
Post a Slack message in #support-urgent with the AI summary and a link to the ticket.
- • Format a clean Slack message
- • Mention the on-duty teammate
- 6Human reviews suggested reply
Suggested reply is a draft, not auto-sent. A human reviews and clicks send.
- • Save draft in Gmail or helpdesk
- • Notify the support agent
- 7Triaged and routed
Every support email is tagged, routed, and ready for a human reply.
Want this adapted to your business?
Copy this recipe to your dashboard and tweak it with the AI assistant.
AI prompts
You are a Shopify support triage assistant. Read the customer email and the recent order context. Return strict JSON: {"category": one of ["refund","shipping","product","account","other"], "urgency": one of ["low","normal","high"], "summary": one short sentence, "suggested_reply": draft 2-3 sentences in a friendly tone}. Email: {{email_body}}. Recent order: {{order_json}}.Human approval recommended
- Always have a human send the final reply, not the AI.
- Review classifications weekly and tune the prompt.
Testing checklist
- Send 5 test emails covering refund, shipping, product, urgent and other.
- Confirm the AI returns valid JSON for each.
- Confirm urgent ones land in Slack with the right info.
- Confirm draft replies appear in the helpdesk for human review.
Error handling
- If OpenAI fails or returns invalid JSON, route the email to a 'manual review' label.
- If Shopify lookup fails, still classify the email and skip the order context.
Common mistakes to avoid
- Auto-sending AI replies without human review.
- Not handling cases where the customer has no Shopify order.
- Forgetting to limit the prompt to a single language.
Frequently asked questions
Need a workflow built around your exact tools?
Generate your own automation recipe with setup steps, AI prompts, platform guidance, and PDF export.